Independent software studio for applied AI and automation

Agentic systems that turn messy workflows into reviewable outputs.

Luminal Works builds lightweight products, prototypes, and internal tools across media, knowledge work, and creative technology.

Work with us See project proof

Current focus

Tools that clarify the work.

01

Agent orchestration and tool-use

Specialist agents, routing logic, MCP-style tool connections, and structured outputs for workflows that need to act, not just answer.

02

Document intelligence

Summaries, citations, extraction, and review layers for long-form inputs where source traceability matters.

03

Content workflows with review

Generation pipelines that produce drafts, scripts, briefs, and manifests with human judgment kept in the loop.

04

Vertical internal tools

Small, practical systems for teams that need a repeatable workflow more than a broad platform.

Project proof

Public work with agentic patterns.

Selected from public repositories and presented as evidence of workflow design, orchestration, retrieval, and generation patterns.

Public repo Multi-agent analysis

Die Hard Script Analyzer

Coordinates specialist analysis agents around screenplay coverage so messy creative inputs become comparable, reviewable notes.

  • Agent orchestration
  • AWS Bedrock
  • Structured review
View repository
Public repo Document workflow

Meeting Transcript Analyzer

Turns long transcripts into grouped questions, summaries, action items, and downstream artifacts for knowledge-work review.

  • Summarization
  • Action extraction
  • Batch workflow
View repository
Public repo Evaluation workflow

BeerJudge

Supports competition judging workflows with structured feedback generation for domain-specific review.

  • Feedback generation
  • Domain review
  • Structured output
View repository
Public repo Content pipeline

Podcast Generator

Automates parts of the research-to-episode workflow while keeping freshness, structure, and editorial review visible.

  • Generation workflow
  • Editorial artifacts
  • Automation
View repository

Redacted work

Private systems summarized safely.

Additional private work includes multi-agent screenplay coverage, enterprise document search, dispatch/audio event extraction, asset-management assistants, trip-planning orchestration, and local-business video generation.

These can be turned into short case studies once sanitized details are approved: domain, inputs, agent or workflow pattern, output artifact, and stack.

How we work

Small systems, clear artifacts, fast iteration.

Clarity

Interfaces, instructions, and outputs should make the next decision easier to review.

Pragmatism

We ship the smallest useful system first, then harden the parts that prove valuable.

Iteration

Prototypes are treated as learning instruments, with evaluation and feedback built in.

Responsible AI

Agentic workflows need guardrails, source visibility, and humans who can inspect the result.

Let's build something useful.

Open for selective consulting, product prototypes, and competition-ready agent workflows.